I have a question with regard to non-parametric bootstrap re-sampling analysis.

I have a sample of n=200 values, and I have performed a bootstrap re-sampling creating 10.000 samples of the same n=200 and then calculated 95%CI, mean, st. deviation.

Can I perform the same bootstrap re-sampling creating 10.000 samples of different sizes (i.e. n=20, n=50, n=100, n=150) bootstrapped from the same original sample with n=200 values?

My objective is to see the impact of sample size (of my original training set) on the 95%CI, mean and st. deviation.

  • $\begingroup$ Have you taken a look at bootstrap theory? E.g., Davison and Hinkley (1997) Bootstrap Methods and Their Application. CUP. $\endgroup$ – utobi Nov 1 '16 at 9:24
  • $\begingroup$ Thanks for your suggestions :) I am new to bootstrap analysis so I might be posting stupid questions, but what is not clear to me is: If I can re-sample with replacement and create a number of bootstrap samples of the same size as my training set, why can't I use the same method (re-sample with replacement) and create bootstrap samples with varying size? Thank you in advance Maria $\endgroup$ – Maria De Leone Nov 1 '16 at 14:09
  • $\begingroup$ Well, you can create samples of varying sizes by resampling from the data with replacement, but that is called bootstrap only when resample size is equal to original sample size. Why do you want other sizes? $\endgroup$ – kjetil b halvorsen Mar 22 '18 at 19:37

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.